lkml.org 
[lkml]   [1998]   [Oct]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [ME TOO] Re: Linux-2.1.125 ... pre-2.2 imminent - SCSI issues

On Sun, 11 Oct 1998, Matthias Andree wrote:

> On Fri, Oct 09, 1998 at 12:08:30PM +0200, Harald Koenig wrote:
> > what about the strange/severe probelms with SCSI tape support right now
> > (see "2.1.x aha1542 scsi tape problems revisited" and just now
> > "Scsi tape - returns ENXIO after an error" threads) ?
> >
> > tapes for backup and data exchange _really_ should be solid in 2.2.0...
>
> This is a bit of an ME TOO, I second this. I was mistaken by the
> assumption that an "old", established product such as the Adaptec
> 1542CF has a reliable support, but here's my experience:

The 1542CF is known to have bad SCSI lines drivers (hardware).
In real life you can compare it with a car that has bad brakes and
bad tyres.

> Adaptec 1542 (terminated), 2 m cable to HP 35480A, active terminator,
> no other devices attached: some major lockups during dump.
>
> Tekram DC-390 U (auto termination, SYM53C875 based), 2 m cable to HP
> 35480A, 1m+anythng cable to Micropolis hard disk, Yamaha CD writer,
> Plextor CD-ROM drive, actively terminated. Actually beyond specs.
> Stress tested fine. Although this is a lucky coincidence and nothing to
> rely on, I mention it to contrast it.

For the same reason you probably donnot drive your family using a
car designed 30 years ago, I suggest you not to use too old SCSI
controllers to drive your SCSI devices.

> There is absolutely no reason why the Adaptec 1542 could lock up the
> entire dump process in state "D" (not even giving timeouts or
> something) and why it need not recover from fixed bus faults (id est:
> have an unterminated SCSI cable with loose end lying around, get bus
> errors, remove the cable => aha1542 still loops aborting/resetting).
> Removing cables from different adaptors such as SYM53c875 gets them
> back up working. Besides, my 1542 disabled itself after locking, and
> needed to be powered-down to get it back up working. IF the aha1542 has
> means to disable itself to avoid damage because of bus faults, the
> driver should be aware of this and in case return error instead of
> locking.

A sophisticated error recovery stuff may allows to be still alive after
some not too serious error, but it is not normal to get errors and when
you get any it is advisable to try to fix the cause as soon as possible
and not to expect error recovery to durably cope with your broken system.

> I am regretfully not the one to debug low-level Kernel scsi drivers, I
> have no low-level Kernel experience, I don't have the Adaptec 1542
> specs either.
>
> IF 2.2 is released without these issues resolved, we'd better mark
> Adaptec 1542 support "EXPERIMENTAL" again.

Support of old devices should be turned to obsolete at the time these
devices become not reliable enough for modern systems. We still can
support minimal drivers for these devices, but should clearly discourage
users to use them on their systems.


Regards,
Gerard.



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:44    [from the cache]
©2003-2011 Jasper Spaans